PII-224 - PREDICTORS OF ALZHEIMER'S DISEASE AND RELATED DEMENTIA (ADRD) IN A HYPERTENSION (HTN) POPULATION WITHIN THE STATE OF FLORIDA USING ELECTRONIC HEALTH RECORD (EHR)-BASED DATA.

E. Jafari¹, S. Smith¹, W. Hogan², G. Smith³, C. McDonough¹; ¹College of Pharmacy, University of Florida, ²Data Science Institute, Medical College of Wisconsin, ³College of Public Health and Health Professions, University of Florida.

Background: HTN is a known modifiable risk factor for ADRD; whether variance in HTN control, antihypertensive (anti-HTN) medication use, and healthcare encounters predict ADRD over time is unknown.

Methods: Validated computable phenotypes were applied to EHR data from the OneFlorida Data Trust (index period 1/1/2013-12/31/2016), to identify HTN phenotypes (controlled HTN, uncontrolled HTN, apparent treatment-resistant HTN [aTRH], and other) and HTN index date. The primary outcome, ADRD, was captured after the HTN index date through 12/31/2019 using validated ICD-9/10 codes. Multiple adjusted stepwise logistic regression models were used to assess the predictability of ADRD with overall controlled and uncontrolled HTN (Model 1), aTRH and other HTN control levels (Model 2), and additional HTN metrics (anti-HTN medications, anti-HTN medication count, visit count, Model 3) in the HTN population.

Results: 101,148 patients were included. Average age was 64 y, with 55% females, 65% White, and 28% Black. 27,257 (27%) were classified as controlled HTN and included well-controlled (n=14,752, BP controlled at >80% of visits) and intermediate-controlled (n=12,505, BP controlled at 50-80% of visits). 29,572 (29.2%) were classified as uncontrolled HTN and included aTRH (n=18,465) and uncontrolled (n=11,107, BP controlled at < 50% of visits). 11,049 (11%) patients were diagnosed with ADRD after the HTN index date (n=2,822, 10.3% controlled HTN patients and n=3,554, 12% uncontrolled HTN patients). Models 1 and 2 showed that uncontrolled BP and aTRH were significant predictors of ADRD (Table 1). When anti-HTN medication variables were included in the regression models, aTRH was no longer significant, and thiazide diuretic use and combination anti-HTN medication preparations were found protective of ADRD (Table 1). However, we observed significant associations between aTRH and anti-HTN medication variables (p>0.0001), suggesting collinearity. Other significant ADRD predictors were: depression [1.52, (1.42 - 1.63)], stroke [1.41, (1.27- 1.58)], brain injury [1.33, (1.21 - 1.47)], and diabetes [1.26, (1.19 - 1.33)].

Conclusion: More complex BP phenotypes, such as aTRH and other correlated variables were stronger predictors of ADRD phenotypes in OneFlorida compared to overall BP control.